The present invention is comprised within the field of mobile telecommunications. More specifically, the invention relates to a system for carrying out a dynamic selection of HSPA+ technology (MIMO or DC, Dual Cell) for mobile terminals able to support MIMO as well as a DC connection.

HSDPA (High Speed Downlink Packet Access) is a packet-based data service in the 3rd generation W-CDMA (Wideband Code Division Multiple Access) systems, which provides high-speed data transmission (with different theoretical peak download rates according to the HSDPA technology step e.g. 7.2/10.8/16.2/21.6/28.8/43.2 Mbps over a 5MHz bandwidth) to support multimedia services.

In order to reach the theoretical peak rates of respectively 28.8 Mbps with 3GPP Release 7 and 43.2 Mbps with Release 8, the MIMO (Multiple Input Multiple Output) feature with respectively 16QAM and 64QAM modulation is used in HSDPA, in which multiple antennas are implemented at both base station (Node B) and mobile terminals (UE: User Equipment).

The basic MIMO feature as standardised in 3GPP Release 7 requires (just) two transmitter antennas (at the node B) and two receiving antennas (at the UE) using a common carrier, as shown in Figure 1, which illustrates a diversity/MIMO system configuration with two RF transmit branches (UTRAN side) and two receive antennas (UE side) which operate at the same frequency, the required baseband processing for MIMO being common for both RF branches at UTRAN and UE side. At UTRAN side, the transmitted data is divided into two data streams and transmitted through the two antennas using the same radio resource (same time i.e. Transmission Time Interval (TTI) and HSDPA codes). The two streams of data are recovered by the UE from the signals received via its two antennas. Clearly then, the MIMO feature needs support in MIMO-enabled terminals as well as in the network.

The patent application US 2007/049218 A1 describes a MIMO based system in which a switching between different communication modes is performed.

In order to deploy MIMO at the node B and to transmit two parallel data streams, two power amplifiers are required per sector (one for each of the two antennas). In order not to limit the use of an entire carrier for MIMO only (5MHz), it is more efficient and practical to use the same carrier as devices that are not MIMO-enabled (e.g. HSDPA legacy terminals) to utilise all available capacity.

MIMO technology is an important step in the evolution of HSDPA, as it provides higher data rates in downlink whilst further improving spectrum efficiency.

A known technique for increasing the capacity of a cellular telecommunications network such as 2G, HSDPA, LTE, WiFi or WiMAX is to bond together two or more carriers. This approach is called carrier or spectrum aggregation (or DC, Dual Cell). MIMO (Multiple-Input Multiple-Output) functionality was standardised as part of the 3GPP Release 7 set of features, whilst Dual Cell (DC) has been standardised in Release 8 for HSDPA operation but only for non-MIMO HSDPA. Therefore a Release 8 DC terminal can either be operated as DC or MIMO but not both simultaneously (however, operation of DC with MIMO will be possible in Release 9), and correspondingly a cell capable of supporting DC and MIMO has to decide which of these two functions to activate for Release 8 UEs (User Equipments) present (i.e. MIMO operation or Dual Cell operation).

The 3GPP TS (Technical Specification) 25.306 defines the HSDPA UE categories. The following table captures the capabilities for UE up to Release 8. HS-DSCH category Max number of HS-DSCH codes received Max number of bits of an HS-DSCH transport block received within a TTI Supported modulations MIMO and Dual cell operation Max Throughput Category 1 5 7298 QPSK, 16QAM (MIMO not supported) 3,6 Category 2 5 7298 3,6 Category 3 5 7298 3,6 Category 4 5 7298 3,6 Category 5 5 7298 3,6 Category 6 5 7298 3,6 Category 7 10 14411 7,2 Category 8 10 14411 7,2 Category 9 15 20251 10,1 Category 10 15 27952 14,0 Category 11 5 3630 QPSK (MIMO or DC not supported) 1,8 Category 12 5 3630 1,8 Category 13 15 35280 QPSK, 16QAM, 64QAM (MIMO or DC not supported) 17,6 Category 14 15 42192 21,1 Category 15 15 23370 QPSK, 16QAM 23,4 Category 16 15 27952 MIMO supported 28,0 Category 17 15 35280 QPSK, 16QAM, 64QAM 17,6 23370 QPSK, 16QAM MIMO supported 23,4 Category 18 15 42192 QPSK, 16QAM, 64QAM 21,1 27952 QPSK, 16QAM MIMO supported 28,0 Category 19 15 35280 QPSK, 16QAM, 64QAM MIMO supported 35,3 Category 20 15 42192 42,2 Category 21 15 23370 Dual Cell QPSK, 16QAM 23,4 Category 22 15 27952 28,0 Category 23 15 35280 Dual Cell QPSK, 16QAM, 64QAM 35,3 Category 24 15 42192 42,2

UE categories 15 to 20 are MIMO capable with up to 42 Mbps of bit rate at physical layer without any physical layer coding -but not Dual Cell-. UE Categories 21 to 24 define the Dual Cell capable UEs that can also be MIMO capable. The UE will signal its category to the network and this will provide an indication to the network if it is DC capable or not -regardless of whether DC is in operation within the cells the UE finds itself in-.

MIMO UEs of Release 8 use only 1 WCDMA carrier of 5 MHz whilst Dual Cell capable UE use 2 contiguous carriers (10 MHz spectrum).

UEs of category 21 to 24 can support both MIMO and DC, it is the network which must decide which of these two functionalities it should operate i.e. configure one but not the other.

In short there will be Node Bs operating with both MIMO and DC active within the same cell. MIMO will be used for UEs that do not support DC; DC will be used by UEs that will support DC only, but for the UE supporting both DC and MIMO capability, an efficient rule has to be established in order to decide which technology should be used for each of these UE when entering the system. These rules have not been specified by 3GPP. It should be noted that in Rel'9 HSPA evolves towards the joint use of Dual Carrier and MIMO for data connections. Following the evolution from these 3GPP capabilities in time, one UTRAN cell could support MIMO (Rel'7) then DC or MIMO on a per user basis (Rel'8) to end up with full DC MIMO capabilities (Rel'9) allowing the joint use of both capabilities for a single user connection.

The present invention overcomes the commented problems by providing a method for automatically assigning to a UE capable of both MIMO and DC (but not simultaneously) the best technology to be used (MIMO or DC) according to the context in terms of radio conditions and traffic. This way the end user would obtain the best possible performance. Thus, the invention increases the data throughput supported by a mobile broadband terminal and increases the capacity of the supporting network.

It is well-known that abbreviations and acronyms are frequently used in the mobile telephony field. Below is a glossary of acronyms/terms used throughout the present specification:

3GPP

The 3rd Generation Partnership Project

CPICH

Common Pilot Channel

CQI

Channel Quality Indicator

DC

Dual Cell

HSPDA

High Speed Downlink Packet Access

KPI

Key Performance Indicators

MAC-d

Medium Access Control- dedicated

MIMO

Multiple-Input Multiple-Output

QAM

Quadrature Amplitude Modulation

QPSK

Quadrature Phase-Shift Keying

RNC

Radio Network Controller

SPI

Scheduling Priority Indicator

SRNC

Serving Radio Network Controller

TTI

Transmission Time Interval

UE

User Equipment

WCDMA

Wideband Code Division Multiple Access

The invention relates to a network element in a cellular telecommunications network for improving network performance according to claim 1, and to a corresponding method according to claim 8. Preferred embodiments of the network element and of the method are defined in the dependent claims. The present invention is defined and limited only by the scope of appended claims 1-13.

A series of drawings which aid in better understanding the invention and which are expressly related with an embodiment of said invention, presented as a non-limiting example thereof, are very briefly described below.

• Figure 1 shows the transmitter and receiver arrangement required for MIMO.
• Figure 2 shows, as an example, a graphic of transmitted TX Power in HS transmission and its standard deviation, used as input parameters in the invention.
• Figure 3 illustrates an example of MIMO gain depending on the CQI of the mobile terminal.

When a UE is capable to support both MIMO and DC technology there is the need to decide which technology is the best to use.

When at least 2 carriers are available and a cell supports the activation of MIMO and DC at the same time, and when an incoming user with a Release 8 UE that can support both MIMO or DC (i.e. not simultaneously) establishes a connection, there is a need to decide if the experience will be optimised by providing MIMO or providing Dual Cell.

In order to compare performance of MIMO and DC with a Release 8 UE, HSDPA performance with 15 codes and 16QAM or 64QAM modulation (whichever modulation is available on the MIMO carrier for the HSDPA transmissions) is taken as a single carrier baseline reference configuration. The performance gain in terms of user throughput of MIMO (due to the higher spectral efficiency) with respect to a 64QAM single carrier reference baseline is around 20-30% in bad to medium radio conditions, and is around 50-60% in good radio conditions. These radio conditions, which can be expressed in terms of CQI, are sent by the UE to the Node B.

The DC technology consists of using 2 carriers simultaneously for a given user, with the capability to use up to 30 HSDPA codes (15 codes per carrier) using QPSK, 16QAM or 64QAM. The spectrum efficiency gain of DC is slightly better than that of 64QAM due to the carrier scheduling gains i.e. having the possibility to select on a TTI basis the carrier which balances best radio conditions and resource availability.

The DC provides great advantages versus the single carrier case since a UE can receive additional throughput using the available capacity on 2 carrier frequencies. Considering f1 as the first carrier frequency used for R99 services and f2 as the carrier frequency used for the HSPA services, the DC gain is highly influenced by the amount of load on f1, by the burstiness of the load (high variance versus low variance), since a bursty load provides several unused TTIs (Transmission Time Intervals) with high available power to be used.

As an example, currently in almost all 3G deployments where 2 carriers are deployed, the first carrier is already assigned for use to R99 voice traffic -hence the load can be such that potential resources available for DC on f1 are more or less scarce depending whether the considered site is urban, suburban, or rural. Furthermore, when sharing a carrier frequency between Rel'99 and HSPA services the maximum HSPA power is sometimes capped in order to not affect the voice quality and related KPIs e.g. call drop rate due to handover failures following the impact of HDSPA power on CPICH Ec/No as perceived by the UE. In this situation, if the load on f1 is quite high, the additional throughput would be limited.

A typical scenario for which the selection of MIMO or DC needs to be decided is the following:

• N carrier frequencies available in the cell.
• MIMO supported in one of the carrier frequency.
• Rel'99 and HSDPA (single carrier non MIMO terminals) also supported.

According to a preferred embodiment of the invention, the UE on a specific radio condition (CQI) will be assigned MIMO when the expected throughput with MIMO (TH_MIMO) is greater than the expected throughput with DC (TH_DC), which can be expressed as: $${\mathit{TH}}_{\mathit{MIMO}}>{\mathit{TH}}_{\mathit{DC}}$$

First step is to select the most adequate carrier frequencies to be used for MIMO and for DC:

• For MIMO, the least loaded carrier frequency enabled for MIMO will be the most adequate, which can be called F_M.
• For DC when there are more than 2 carriers, then the 2 least loaded carrier frequencies supporting DC will be selected. The least loaded carrier frequency will be F_M; the other one will be named F_DC2.

Then, expected throughput with DC (TH_DC) can be expressed as the expected throughput in F_M without MIMO (TH(F_M)_{NO_MIMO}) plus the additional expected throughput the UE can get by also using F_DC2 as part of DC (ΔTH(F_DC2)_DC). Then, TH_DC can be expressed as: $${\mathit{TH}}_{\mathit{DC}}=\mathit{TH}{\left({\mathit{F}}_{\mathit{M}}\right)}_{\mathit{NO\_MIMO}}+\mathit{\Delta TH}{\left({\mathit{F}}_{\mathit{DC}\mathit{2}}\right)}_{\mathit{DC}}$$

Therefore, the UE will be assigned MIMO when the expected throughput in F_M with MIMO (TH(F_M)_MIMO) is greater than TH_DC as previously expressed: $${\mathit{TH}\left(F_M\right)}_{\mathit{MIMO}}>\mathit{TH}{\left({\mathit{F}}_{\mathit{M}}\right)}_{\mathit{NO\_MIMO}}+\mathit{\Delta TH}{\left({\mathit{F}}_{\mathit{DC}\mathit{2}}\right)}_{\mathit{DC}}$$

The expected throughput in F_M with MIMO TH(F_M)_MIMO can be expressed as: $${\mathit{TH}\left(F_M\right)}_{\mathit{MIMO}}=\mathrm{expected\; tput\; in}{\mathrm{F}}_{\mathrm{M}}\mathrm{without\; MIMO}*\left(\mathrm{1}+\mathrm{MIMO\; Gain}\right)$$

Since the expected throughput in F_M without MIMO can be expressed as TH(F_M)_{NO_MIMO}, and MIMO Gain as G_MIMO, the previous formula is expressed as: $${\mathit{TH}\left(F_M\right)}_{\mathit{MIMO}}=\mathit{TH}{\left({\mathit{F}}_{\mathit{M}}\right)}_{\mathit{NO\_MIMO}}\cdot \left(1+{\mathrm{G}}_{\mathrm{MIMO}}\right)$$

Therefore, using MIMO is preferably over DC if: $${\mathit{TH}\left(F_M\right)}_{\mathit{NO\_MIMO}}\cdot \left(1+G_{\mathit{MIMO}}\right)>\mathit{TH}{\left({\mathit{F}}_{\mathit{M}}\right)}_{\mathit{NO\_MIMO}}+\mathit{\Delta TH}{\left({\mathit{F}}_{\mathit{DC}\mathit{2}}\right)}_{\mathit{DC}}$$

Each of the items above are functions of CQI and occupied TX Power in F_M by other traffic as well as by the traffic characteristics, that, through the scheduler implementation, will turn into specific variance of the transmitted power.

The condition above is evaluated as a function on the CQI experienced by the UE on F_M and F_DC2. TH(F_M)_{NO_MIMO}(CQI) can be modelled as a function of the available resources for HSPA transmission in the cell, i.e. TX power available for HS (denoted "Pfree"), Codes. Thus, $${\mathit{TH}\left(F_M\right)}_{\mathit{MIMO}}=\mathrm{MAXRateHS}\left(\mathrm{CQI},\mathrm{Pfree\; for\; HS},\mathrm{code\; available\; for\; HS}\right)$$ where MAXRateHS denotes the maximum throughput that would be achieved with HSDPA for a given level of CQI, a given power available for HSDPA and a given number of HSDPA codes available.

From simulations it has been observed that the gain of Dual Cell respect to Single Carrier is linked to the likelihood of there being free resources for transmission in F_DC2 since the scheduler can take advantage of any free TTI or free code per TTI on F_DC2 to transmit packets of a DC UE.

The expected throughput in a DC transmission can be expressed as the average amount of throughput that is achievable on both carriers: $${\mathit{TH}}_{\mathit{DC}}=\mathrm{MAXRateHS}\left(\mathrm{CQI\; in}{\mathrm{F}}_{\mathrm{M}},\mathrm{Pfree\; for\; HS\; in}{\mathrm{F}}_{\mathrm{M}},\mathrm{code\; available\; for\; HS\; in}{\mathrm{F}}_{\mathrm{M}}\right)+\mathrm{MAXRateHS}\left(\mathrm{CQI\; in}{\mathrm{F}}_{\mathrm{DC}\mathrm{2}},\mathrm{Pfree\; for\; HS\; in}{\mathrm{F}}_{\mathrm{DC}\mathrm{2}},\mathrm{code\; available\; for\; HS\; in}{\mathrm{F}}_{\mathrm{DC}\mathrm{2}}\right)\mathrm{.}$$

The condition for selecting MIMO preferably to DC on next TTI would therefore be: $$\begin{array}{l}\mathrm{MAXRateHS}\left(\mathrm{CQI\; in}{\mathrm{F}}_{\mathrm{M}},\mathrm{Pfree\; for\; HS\; in}{\mathrm{F}}_{\mathrm{M}},\mathrm{code\; available\; for\; HS\; in}{\mathrm{F}}_{\mathrm{M}}\right)\cdot \left(1+G_{\mathit{MIMO}}\left(\mathrm{CQI}\right)\right)>\\ \mathrm{MAXRateHS}\left(\mathrm{CQI\; in}{\mathrm{F}}_{\mathrm{M}},\mathrm{Pfree\; for\; HS\; in}{\mathrm{F}}_{\mathrm{M}},\mathrm{code\; available\; for\; HS\; in}{\mathrm{F}}_{\mathrm{M}}\right)+\mathrm{MAXRateHS}\left(\mathrm{CQI\; in}{\mathrm{F}}_{\mathrm{DC}\mathrm{2}},\mathrm{Pfree\; for\; HS\; in}{\mathrm{F}}_{\mathrm{DC}\mathrm{2}},\mathrm{code\; available\; for\; HS\; in}{\mathrm{F}}_{\mathrm{DC}\mathrm{2}}\right)\end{array}$$

The previous formula can be simplified. Whilst the formula above would allow a perfect choice on a TTI basis, there is the need to come up with a calculation that does not need to be executed on each TTI, and that, based on the average load and HSDPA usage in the cells, is able to predict the future cell conditions and come up with a proper decision to be applied at call set-up or sometimes during the call.

In a first formula simplification, it can be assumed that MAXRateHS is directly proportional to the amount of power available for HS as well as to the average number of codes available for HSDPA on average (i.e. taking away the codes used for Rel'99 traffic and by the existing HSDPA transmission). Then: $$\mathit{MAXRateHS}=\mathit{Alfa}\left(\mathit{CQI}\right)*\left(\mathit{Pfree\; for\; HS}\right)*\left(\mathit{free\; HS\; codes}/\mathit{15}\right)$$ where Alfa(CQI) is a factor which represents the throughput that would be achieved with 15 codes and full HSDPA power available for a given CQI. This can be simulated or obtained through the experimental tests and can be expressed as a function or a table of values from 0 to 30 to cover the full the CQI range. The CQI is the Channel Quality Indicator and it is a parameter defined in the 3GPP standards (TS 25.214). It should be noted that the CQI definition can vary according to the HSDPA UE category, so the Alfa(CQI) will be different depending on the UE category that is under evaluation. Then there will be different tables for different UE categories.

The previous formula would become: $$\begin{array}{l}\mathit{Alfa}\left(\mathit{CQI\; in}{\mathit{F}}_{\mathit{M}}\right)*\mathit{Pfree\; for\; HS}\left({\mathit{F}}_{\mathit{M}}\right)*\left(\mathit{average\; free\; HS\; codes\; in}{\mathit{F}}_{\mathit{M}}/\mathit{15}\right)*\left(\mathit{1}+{\mathit{G}}_{\mathit{MIMO}}\left(\mathit{CQi}\right)\right)>\\ \mathit{Alfa}\left(\mathit{CQI\; in}{\mathit{F}}_{\mathit{M}}\right)*\mathit{Pfree\; for\; HS}\left({\mathit{F}}_{\mathit{M}}\right)*\left(\mathit{average\; free\; HS\; codes\; in}{\mathit{F}}_{\mathit{M}}/\mathit{15}\right)+\mathit{Alfa}\left(\mathit{CQI\; in}{\mathit{F}}_{\mathit{DC}\mathit{2}}\right)*\mathit{Pfree\; for\; HS}\left({\mathit{F}}_{\mathit{DC}\mathit{2}}\right)*\left(\mathit{average\; free\; HS\; codes\; in}{\mathit{F}}_{\mathit{DC}\mathit{2}}/\mathit{15}\right)\end{array}$$

Therefore: $$\begin{array}{l}\mathit{Alfa}\left(\mathit{CQI\; in}{\mathit{F}}_{\mathit{M}}\right)*\mathit{Pfree\; for\; HS\; in}{\mathit{F}}_{\mathit{M}}*\left(\mathit{average\; free\; HS\; codes\; in}{\mathit{F}}_{\mathit{M}}/\mathit{15}\right)*{\mathit{G}}_{\mathit{MIMO}}\left(\mathit{CQI}{F}_M\right)>\\ \mathit{Alfa}\left(\mathit{CQI\; in}{\mathit{F}}_{\mathit{DC}\mathit{2}}\right)*\mathit{Pfree\; for\; HS\; in}{\mathit{F}}_{\mathit{DC}\mathit{2}}*\left(\mathit{average\; free\; HS\; codes\; in}{\mathit{F}}_{\mathit{DC}\mathit{2}}/\mathit{15}\right)\end{array}$$

In a second formula simplification, it can be assumed that the average CQI in the 2 carriers will be the same for the UE, i.e. Alfa (CQI in F_M) = Alfa (CQI in F_DC2).

Thus, MIMO will be selected if: $$\mathit{Pfree\; for\; HS\; in}{\mathit{F}}_{\mathit{M}}*\left(\mathit{average\; free\; HS\; codes\; in}{\mathit{F}}_{\mathit{M}}/\mathit{15}\right)*{\mathit{G}}_{\mathit{MIMO}}\left(\mathit{CQI}{\mathit{F}}_{\mathit{M}}\right)>\mathit{Pfree\; for\; HS\; in}{\mathit{F}}_{\mathit{DC}\mathit{2}}*\left(\mathit{average\; free\; HS\; codes\; in}{\mathit{F}}_{\mathit{DC}\mathit{2}}/\mathit{15}\right)$$

From a practical point of view, the RNC at call set-up will decide if the UE will be using DC or MIMO. Any of the formula above, or any other similar or equivalent formula, has to therefore be applied based on the info available at the RNC at call set-up.

The formulas above can also be executed later on during the call after an amount of packets has been transmitted in order to better evaluate the radio conditions (e.g. the CQI) and re-evaluate resource availability.

Then the preferred procedure would be to perform a first selection at call setup based on the above-mentioned formulas, and then during the call evaluate the conditions and do the reconfiguration if needed.

At call setup CQI information is likely to not be available (unless HSPA was used in cell_FACH state), hence an assumption can be made on a CQI level based on other information available such as CPICH Ec/NO, CPICH RSCP, and Transmitted Carrier Power which allows to have an estimate of the radio conditions. One solution is to define an RNC level parameter representing the estimated radio conditions from the user with e.g. 3 basic levels (bad, medium, good) and use this parameter as an input to the formulas by assigning a CQI value to each radio condition level (for example: 15 for bad, 21 for medium, 26 for good). With this approach one can either use this parameter in order to derive the most efficient technology to use at call setup or derive the best value for this parameter based on current estimates available at call setup.

Another alternative at call setup is to make a selection between MIMO or DC based on a default RNC level parameter, and then perform a selection of the most appropriate technology based on the above formulas after call setup.

In order to perform a (re)-selection of the most appropriate technology between MIMO and DC after call setup, two types of trigger can be used:

1. 1) Timers: perform a re-evaluation after a certain connection time and/or periodically e.g. every 10 minutes.
2. 2) Resource availability on the carrier frequencies: the RNC monitors the available power and codes for HSDPA and decides to force a re-evaluation. This can be implemented in an event triggered way to save signalling

The probability to get a free TTI or an amount of codes in Downlink during HSDPA transmission could be calculated through the standard deviation of the transmitted TX Power. The standard deviation of the HSDPA transmitted power (as shown in the example of Figure 2) together with the average transmission power could be exploited in the above formulas. For example: $$\mathit{PFree\; for\; HS}=\mathit{Min}\left(\mathit{Average\; free\; power}+½*\mathit{Power\; Std\; Deviation},\mathit{MaxHSPower}\right)$$ where MaxHSPower is the maximum power available for HS). This allows to factor in the burstiness of the traffic when selecting between DC and MIMO

The MIMO gain (versus the single carrier HSPA baseline) could be given as a single value set by a parameter or a reference table per CQI. In this case, the decision should be taken knowing what the average CQI is at the time of the evaluation. Maybe it could be established with DC or MIMO first and then do a reconfiguration to the selected technology. For example the MIMO gain as a function of the CQI, can be approximated as represented in Figure 3.

There are other possible criteria to select between MIMO or DC at call setup or during the call:

• ∘ Number of users currently using MIMO or DC in the system.
• ∘ Traffic characteristic of a specific user

The selection between MIMO and DC operation can be based upon the number of users present in the cell(s) at a point in time e.g. balancing the number of calls in DC and the number of MIMO calls. For example, in case there are 4 MIMO calls and 2 DC calls, the next one will be DC. This is the simplest solution, easy to implement and enables performance optimisation if MIMO is selected, and an inter-carrier balancing and frequency utilisation efficiency if the DC feature is selected.

The selection between MIMO and DC operation can also be based upon the traffic characteristic of the user. In a typical 2 carrier 3G deployment where F_DC2 is used primarily for R99 services, DC benefits become apparent where most of the R99 traffic is non-voice, because (non HSPA) PS traffic is typically quite bursty in nature i.e. it is then possible for DC to take advantage and utilise the "free" TTIs.

It is not enough to attempt to re-use existing QoS classes to determine DC vs MIMO selection at establishment of those connections as this will not give the level of granularity needed over a specific point in time.

At user level in order to optimise capacity and user performance, the burstiness of the data traffic from the user could be categorised (continuous/bursty) at SRNC level by monitoring the MAC-d (Medium Access Control- dedicated) flow control. According to the activity factor estimated for the user (periodicity of data transfer/request to/from Node B), the SRNC will favour the use of MIMO for continuous traffic and DC for bursty traffic.

For example, the continuous data traffic can be detected with the number of bytes received during a window of time defined: minimum of cumulative received bytes (CRB). Value CRB will be also the minimum data size it wants to be considered as long continuous traffic.

At call setup as mentioned above a parameter can be used to start the data call as MIMO or DC by default, afterwards the categorisation of data traffic for the user is updated during the connection.

Both cell level and user level criteria can be used in the selection process between MIMO and DC by appropriate selection of thresholds. For example for continuous traffic the cell level load based algorithm can be used and for bursty traffic DC would have higher priority. This bursty traffic is detected based on the activity factor. The activity factor is the percentage of the time in which the user is active. If the activity factor is very low means that it transmit in a bursty way and DC is more suitable in this case.

QoS priority of the user can be taken also into account for the DC or MIMO decision. For example with a direct mapping of the priority (SPI, namely Scheduling Priority Indicator) into the selected option, which could be combined with the previous algorithm. The next table is an example: SPI MIMO / DC choice 1 Optimum throughput algorithm 2 Always DC ... 15 Always MIMO

These criteria can be further expanded in order to cover the following points:

• when there are many free resources in F_DC2, the MIMO/DC choice follows the optimum throughput algorithm for all the users.
• when the amount of free resources in F_DC2 becomes lower than a threshold, then only the highest SPIs are given the chance to maximize the throughput occupying F_DC2 resources with DC, whilst all the other SPIs will by default occupying resources on F_M only, using MIMO.